[LLVMdev] Is IR VM does make sense since llvm3 ?

[Dan Gohman]

On Jul 18, 2012, at 2:39 PM, Pierre P <ploploplop123 at gmail.com> wrote:

> Hi llvm list !
> 
> Everything is in the question.
> I've read this discussion on the mailinglist  [LLVMdev] LLVM IR is a compiler IR. 
> But since llvm3 and type system rewrite, is it a good idea to rethink about a VM wich could run the IR bytecode directly?

The type system changes you mention changed the way struct types
are named and uniqued, but structs are still just structs. This
change doesn't really make LLVM IR higher-level in any way that
would significantly affect the issues discussed in that thread.

> 
> llvm has differents bytecode from low level, to more hight level... So do you see some interest to have this kind of VM for one of this bytcode ?
> Is it hight level enougth like java byte code ?

Java bytecode remains much much higher-level than LLVM IR; this
hasn't significantly changed.

> 
> My second obvious question is about the bycode format, is it stable enought to concider using it as an 'archive' source/byte code ? 

Quite a few people are interested in keeping the bitcode format
stable these days, so it will probably remain fairly stable for the
foreseeable future.

That said, as far as I'm aware all of the issues discussed in the
"LLVM IR is a compiler IR" thread are at least as relevant today as
they were then.

Dan